Thedwick

I just got in my inbox a message about the Paas cloud platform formerly known as Makara. Redhat bought the company and has since rebranded it to OpenShift Flex. Looks like a cool offering: super-easy deployment of PHP, Python or Rails apps, or a more custom but complicated deployment of J2EE apps and higher-powered sites. Can’t wait to have an excuse to use it.

The Netflix tech blog has a nice post about their top 5 lessons learned from porting to AWS cloud. They list lessons that I’ve seen frequently in such things, especially “learn to deal with component failure” and “latency is variable” but the most striking thing I saw in there was the Chaos Monkey.

They only mentioned it briefly, but I thought it was both stunning and brilliant. They actually have a process in their (production? wasn’t clear) environment that wanders around randomly killing things, just to keep them honest and make sure failure really is tolerated. What a stroke of genius! Presumably, the Chaos Monkey is scheduled to only do its job during times when it’s convenient to deal with the fallout (ie. not at 3am, nor on Thanksgiving).

This is similar to running business continuity tests. At one of my old companies, we had a philosophy that if the continuity plan hadn’t actually been tested by failing over for realsies in the production environment then it wasn’t really a plan at all–it was just a piece of paper. Doing these tests was incredibly painful and risky, but it allowed us to learn our lessons at convenient times with the least risk to the business (usually weekends ) instead of at peak times. And learn lessons we did. Every. Single. Time.

File this in the “is this 2010 or 1980?” department. I’m working on a financial app in Google App Engine and discovered that GAE doesn’t support precision decimals! How does something like that happen in a modern framework? It’s as if the last 30 years of computing never happened.

I had the chance to check out Makara, a cloud-based platform as a service today. I was impressed; they’ve managed to make deploying scalable Java applications easy, which is no small feat. Anyone who has tried to set up a clustered Java web app can attest that it’s no fun and not an experience you want to repeat. But these guys seem to have figured it out, and they certainly impressed Red Hat enough to want to acquire them.

Plus, their model is so much better than other PaaS offerings like Google App Engine. With Makara, you’re not stuck on one infrastructure. You can deploy (in theory at least) to anything. I’d be interested to know if anyone has actually done that in practice, though.

I honestly don’t believe there is any collective, purposeful, evil intent at Oracle when it comes to Java.   But their actions since the merger with Sun sure make it seem that way.  Oracle is killing Java, one cut at a time.

The first big shot across the bow was Oracle suing Google over Android.  Even if Google did something wrong legally, it was still a stupid move.  After many other companies failed to get Java to thrive on phones (it’s been there for years, it just was clunky and had no developer community), Google finally succeeded with the first plausible challenger to the iPhone.  And, on top of it, the Android community is much more open which attracts even more developers.  So a thriving Android community can only be a good thing for Oracle, even if they don’t get licensing revenue from the runtime.  But the suit demonstrates that Oracle just doesn’t get it.

Then, there was JavaOne.  I covered it for InfoQ, but since we try to stay pretty neutral I didn’t say what I really thought: it was awful.  I’ve never been a fan of JavaOne because it’s too big, too much of a hassle, and too full of vendors.  But this year was even worse.  Adding it to the rest of the Oracle conference took it from merely too big to an absolute zoo.  And, it watered down the Java bit to just another Oracle add-on.  The biggest display in Moscone center, the one that took up the entire entry way when you came in, was about Larry’s yacht, and not technology.  Apollo Ohno was a speaker, as if any of us cared.  And the place was absolutely buzzing with “suits”.  The whole conference had Larry’s ego all over it.

Then James Gosling left, which isn’t a material impact on Java, but it is a symbolic one.

And now the Hudson flap is the last straw for me.  Here, Oracle took a fantastic, thriving open source project that doesn’t produce any revenue, and is squashing it over a pidily infrastructure tiff.  And they’re doing it in the worst way, by saying, basically, “I know this Java.net infrastructure is behind the curve, but we dictate that you should just suck it and use it anyway because we own the trademark.  So there.”  I can’t think of a faster way to kill a community that’s contributing freely.

Just get a clue please Oracle.  I’ve invested a lot of time in Java and would hate have to walk away from it completely.  As it is, I’m already steering greenfield stuff towards Ruby.  Maybe I should follow Nikita’s advice and learn Scala.

There’s a little-known but lucrative prize out there for developers that make software that makes the world better: the Pizzigati Prize for Software in the Public Interest. Good to know sometimes the good guys win, too.

So ever since I started writing for InfoQ, I’ve had a hard time keeping this blog up, too. But I promise I’ll hit the balance and manage to post to both. And soon.

In the mean time, I’m going to post highlights to my InfoQ stories here as well, just to keep things lively.

This week, I wrote a quick post about Spring Social, a new API from SpringSource to integrate Java applications with social networking sites. It’s pretty nifty, and using it feels a lot like using trusty old JdbcTemplate.